Flood modelling and prediction using hydrologic similarity – driven parameter regionalization in ungauged catchments

Flood forecasting in ungauged catchments with scarce hydrological data remains challenging, as traditional parameter regionalization overlooks hydrological heterogeneity, leading to poor peak flow and runoff process prediction. This study proposed a mechanism-based differentiated parameter regionalization framework, decoupling runoff generation and concentration processes via hydrological similarity assessment and parameter sensitivity analysis. Sensitivity analysis reveals that runoff generation parameters are highly sensitive to peak flow, while runoff concentration parameters de-pend on topographic characteristics. Accordingly, runoff generation parameters are transferred from hydrologically sim-ilar reference basins, and runoff concentration parameters are calculated using geographic information system-derived spatial data. A multi-dimensional similarity evaluation framework – integrating static underlying surface features, dy-namic hydrological responses, and process-based mechanisms – addresses hydrological heterogeneity, and thus over-come the uniform parameterization strategy limitation of conventional methods. Application in the middle and upper Chu River Basin shows improved simulation accuracy: the Nash-Sutcliffe Efficiency (NSE) increased by an average of 0.08. The hydrological-hydrodynamic coupled model enables real-time parameter updates using geographic information system and remote sensing data, enhancing dynamic simulation of flood dynamics in ungauged sub-basins. This research advances parameter regionalization by explicitly linking it to hydrological mechanisms, providing a scalable framework for flood forecasting in ungauged catchments worldwide.